Teams preparing for an inspection tend to prepare the wrong artefact. They assemble the current state — every certificate, in date, in one folder. That is necessary and it is not what the difficult questions are about.
The difficult questions are about the past, and they have a recognisable shape.
How did you know?
Not whether the certificate is valid today, but what the process is that would have caught it if it were not. A screenshot of a calendar reminder is a weak answer. A record showing that the system computed a due date, raised a notification against a named owner, and that the owner acknowledged it, is a strong one.
This is why acknowledgement is a first-class concept rather than a nicety. Having emailed someone demonstrates that a job ran. It does not demonstrate that anybody read it, and an inspector is entitled to notice the difference.
What changed, and who changed it?
A calibration interval that moved from six months to twelve is a legitimate engineering decision or a shortcut taken under deadline pressure, and the record has to be able to tell them apart. That means the history needs the before value, the after value, the person, the timestamp, and ideally the reason.
If the history can be edited, none of it counts. This is the practical reason an audit log has to be append-only at the privilege level rather than by convention: the question is not whether you would rewrite it, but whether you could.
Show me the one that failed
The best auditors go looking for the failed inspection, the conditional pass, the out-of-tolerance result. A register that only records successes is not reassuring — it reads as a register that is not being used honestly.
- Record passed, failed and conditional as equal outcomes, not as an exception path.
- Keep as-found and as-left values, because the gap between them is the finding.
- Link the corrective action to the result it came from, so the chain is one hop.
- Never let a failure be quietly replaced by the passing re-test that followed it.
Can you hand me this?
Eventually the request becomes an export: this asset, this date range, everything. If producing it involves an engineer, a query and an afternoon, the system has not actually made you audit-ready — it has moved the work.
An export you can only get by asking your vendor is not an audit trail you control.
The same export should be available to you on demand, in a format you can read without us, covering every table and the attached documents. It is also, not coincidentally, how you leave. A register you cannot walk out of is a hostage rather than a product.